Add options to warp images to target space/resolution

[tsalo]

With antspyx, we can probably apply transforms to standard spaces using pure Python. I'm opening this as a discussion instead of an issue because I think it will require more back-and-forth than a typical feature request, as it opens the door to expanding tedana's scope in possibly controversial ways.

I have been against features like this in the past, since I think that they are better deferred to something like fMRIPost-tedana (a BIDS App I've been planning to work on for a few years, but which hasn't magically materialized). However, I do think it would be reasonable to limit transforms to figures, but not for image outputs.

I think this would be particularly useful for carpet plots, since we could have the voxels organized by tissue type, and component plots, where we'd be better able to localize the high and low weights. The one drawback I can foresee with the component plots is that multiband slice leakage artifacts would be harder to visually identify.

Something like this:

tedana \
    ... \
    --xfms from-T1w_to-MNI152NLin6Asym_mode-image_xfm.h5 from-boldref_to-T1w_mode-image_xfm.txt \
    --reference tpl-MNI152NLin6Asym_res-02_T1w.nii.gz \
    --dseg tpl-MNI152NLin6Asym_res-02_desc-carpet_dseg.nii.gz

@handwerkerd @eurunuela WDYT?

[eurunuela]

I agree with you on that we should limit these to plotting functions. However, I wonder how much computing time it would take and whether it adds enough value to make the extra computing cost worthwhile.

[tsalo]

Opus threw together a branch that works with fMRIPrep transforms very quickly, so I have had a chance to test out the proposal, but I haven't done any benchmarking. I can do that soon. The feature is available in https://github.com/tsalo/tedana/tree/apply-transforms, if you want to try playing with it.

I agree that it's an open question where this would make manual classification/QC easier enough to be worthwhile.